You are on page 1of 11

Nuclear Engineering and Design 245 (2012) 161171

Contents lists available at SciVerse ScienceDirect

Nuclear Engineering and Design


journal homepage: www.elsevier.com/locate/nucengdes

Validation of a FLUENT CFD model for hydrogen distribution in a containment


D.C. Visser , M. Houkema 1 , N.B. Siccama, E.M.J. Komen
Nuclear Research and Consultancy Group (NRG), Westerduinweg 3, 1755 ZG, Petten, The Netherlands

a r t i c l e

i n f o

Article history:
Received 2 November 2011
Received in revised form 27 January 2012
Accepted 30 January 2012
Keywords:
Hydrogen stratication
Severe accident
CFD
Containment thermal hydraulics

a b s t r a c t
Hydrogen may be released into the containment atmosphere of a nuclear power plant during a severe
accident. Locally, high hydrogen concentrations may be reached that can possibly cause fast deagration
or even detonation and put the integrity of the containment at risk. The distribution and mixing of
hydrogen is, therefore, an important safety issue for nuclear power plants.
Computational uid dynamics (CFD) codes can be applied to predict the hydrogen distribution in the
containment within the course of a hypothetical severe accident and get an estimate of the local hydrogen
concentration in the various zones of the containment. In this way the risk associated with the hydrogen
safety issue can be determined, and safety related measurements and procedures could be assessed. In
order to further validate the CFD containment model of NRG in the context of hydrogen distribution
in the containment of a nuclear power plant, the HM-2 test performed in the German THAI (thermalhydraulics, hydrogen, aerosols and iodine) facility is selected. In the rst phase of the HM-2 test a stratied
hydrogen-rich light gas layer was established in the upper part of the THAI containment. In the second
phase steam was injected at a lower position. This induced a rising plume that gradually dissolved the
stratied hydrogen-rich layer from below. Phenomena that are expected in severe accidents, like natural
convection, turbulent mixing, condensation, heat transfer and distribution in different compartments,
are simulated in this hypothetical severe accident scenario.
The hydrogen distribution and associated physical phenomena monitored during the HM-2 test are
predicted well by the CFD containment model. Sensitivity analyses demonstrated that a mesh resolution
of 45 mm in the bulk and 15 mm near the walls is sufciently small to adequately model the hydrogen
distribution and dissolution processes in the THAI HM-2 test. These analyses also showed that wall
functions could be applied. Sensitivity analyses on the effect of the turbulence model and turbulence
settings revealed that it is important to take the effect of buoyancy on the turbulent kinetic energy into
account. When this effect of buoyancy is included, the results of the standard k- turbulence model and
SST k- turbulence model are similar and agree well with experiment. The outcome of these sensitivity
analyses can be used as input for setting up the guidelines on the application of CFD for containment
issues.
2012 Elsevier B.V. All rights reserved.

1. Introduction
During a severe accident in water-cooled reactors, large quantities of hydrogen and steam can be released into the containment.
The hydrogen, generated as a result of core degradation and oxidation, can form a combustible gas mixture with the oxygen present
in the containment atmosphere. Unintended ignition of this mixture can initiate a combustion process, which may damage relevant
safety systems and challenge the integrity of the containment. In
the worst-case, the safety function of the containment can get lost.

Corresponding author. Tel.: +31 224 56 4193; fax: +31 224 56 8490.
E-mail address: visser@nrg.eu (D.C. Visser).
1
Present address: Energy research Centre of the Netherlands (ECN), Petten,
The Netherlands.
0029-5493/$ see front matter 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2012.01.025

The potential danger of hydrogen was rst realized after the


Three Mile Island accident in 1979, where a large quantity of hydrogen was released into the containment and started burning. Since
then, many efforts have been taken to mitigate and/or reduce the
potential risk of hydrogen. For instance, by installation of hydrogen
recombiners that convert hydrogen to steam. The recent hydrogen
explosions during the Fukushima Daiichi accident in March 2011
showed, however, that the control and mitigation of the hydrogen
risk is still a key safety issue for nuclear power plants.
In order to assess the potential risk of hydrogen and the effectiveness of the mitigation systems installed in the containment, it
is necessary to predict the hydrogen concentration in the containment during a severe accident. Two thermal-hydraulic approaches
can be used for this prediction (SOAR, 1999): the lumped parameter
(LP) and the computational uid dynamics (CFD) approach. The LP
codes are of practical use because they are able to give a quick estimate on the hydrogen distribution during a severe accident. The

162

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

Nomenclature
Gb
g
k
Prt
Re

buoyancy effect source term (kg/m s3 )


gravitational acceleration (m/s2 )
turbulent kinetic energy (m2 /s2 )
turbulent Prandtl number ()
Reynolds number ()

Greek symbols

turbulent dissipation rate (m2 /s3 )



density (kg/m3 )
turbulent viscosity (kg/m s)
t

specic dissipation rate (1/s)

high resolution CFD codes provide detailed information on local


phenomena and concentrations. The OECD launched the international benchmark exercise ISP47 in order to assess the capabilities
of both approaches for containment analyses (Allelein et al., 2007).
A combined use of LP and CFD was recommended, where LP will
act as the main workhorse and CFD will be used for more detailed
analyses with high (local) resolution.
The 3D special purpose CFD codes GOTHIC, GASFLOW and
TONUS were the rst CFD codes designed for containment analyses. GOTHIC is an EPRI-sponsored code that can be used for
either lumped-parameter computations or for more detailed multidimensional analysis (Andreani and Paladino, 2010). GASFLOW
is a joint development of Forschungszentrum Karlsruhe and Los
Alamos National Laboratory for the simulation of steam/hydrogen
distribution and combustion in complex nuclear reactor containment geometries (GASFLOW, 2011). TONUS is the French
in-house hydrogen risk analysis code developed by CEA and IRSN
(Kudriakov et al., 2008). Compared to more recent commercial
CFD codes like CFX and FLUENT, these early codes are limited in
their meshing capabilities, turbulence modeling and multiprocessor parallel performance. Therefore the interest and use of the
new, general purpose commercial codes for containment analyses increased substantially over the last years (Heitsch et al., 2010;
Prabhudharwadkar et al., 2011). The GOTHIC, GASFLOW and TONUS
codes are, however, still readily employed because of their specic models (e.g. steam condensation and hydrogen mitigation) and
extensive validation.
Although the CFD codes have proven to be a powerful tool,
extensive validation and clear guidelines are necessary before these
codes can be reliably used for real plant analyses. In the present
work, the containment model developed by NRG in the commercial
CFD code FLUENT is further validated using the well instrumented
and well dened THAI HM-2 test. This test simulates the relevant
physical phenomena involved in the context of hydrogen distribution in a large, multi-compartment containment under severe
accident conditions (Kanzleiter and Fischer, 2008).
The international benchmark exercise ISP47 (Allelein et al.,
2007) as well as the HM-2 benchmark exercise within the
OECD-NEA THAI project (Schwarz et al., 2010) showed a strong
user-dependence, which demonstrates the importance of setting
up and applying best practice guidelines (BPGs) specic for containment applications. To make a start with the development of
such BPGs, sensitivity analyses are performed in the present work
to analyse the effect of mesh resolution, near-wall treatment,
turbulence modeling and turbulence settings on the hydrogen distribution in a containment.
The work presented here is part of NRGs program on the long
term development and validation of a reliable and complete CFD
containment model for hydrogen distribution and combustion,

including mitigation systems such as recombiners, sprays, and condensers.


2. THAI HM-2 experiment
The THAI facility is operated by Becker Technologies in
Eschborn, Germany. The objective of the HM (hydrogen mixing) test
series within the OECD-NEA THAI project was to study hydrogen
mixing and distribution in a large, multi-compartment containment. Test HM-1 was performed with inert helium gas and test
HM-2 with hydrogen gas. A detailed description and comparison
of the HM tests is given by Gupta et al. (2010). The conguration of
the THAI facility, instrumentation and test conditions of the HM-2
experiment are specied by Kanzleiter and Fischer (2008).
Fig. 1 shows the THAI vessel with the internal structures as
employed in the HM tests. The THAI vessel is a cylindrical containment with a height of 9.2 m, a diameter of 3.2 m and a total
volume of 60 m3 . The HM test setup facilitates the study of hydrogen mixing in a multi-compartment facility: the vessel contains an
inner cylinder and four condensate trays that divide the vessel into
a base, a cylinder, an annulus and a dome region. These different
compartments in the THAI vessel are indicated in the schematic
drawing in Fig. 1. The inner cylinder with a height of 4 m and a
diameter of 1.4 m is open at both ends. The condensate trays at
an elevation of 4 m from the bottom of the vessel block 2/3 of
the cross-sectional area in the annulus. Various instrumentation
devices are installed at different locations in the vessel for measuring the hydrogen concentration, temperature, pressure and ow
velocity.
At the start of the HM-2 test, the vessel atmosphere consists
of 98 vol% nitrogen gas, 1 vol% oxygen and 1 vol% steam at ambient conditions (1 bar, 21 C). The HM-2 test can be divided in two
phases;
Phase-1: hydrogen/steam injection and formation of a stable
stratied hydrogen-rich gas layer in the upper part of the vessel
(04300 s).
Phase-2: steam injection, dissolution of the stratied hydrogenrich gas layer and mixing of the atmosphere in the vessel
(43006860 s).
In phase-1, a mixture of hydrogen (0.3 g/s) and saturated steam
(0.24 g/s) is injected in upward direction into the annulus from a
circular pipe of 28.5 mm diameter at an elevation of 4.8 m. The average injection temperature during this phase is 45 C. At the end of
phase-1, from 4200 to 4300 s, there is no injection. In phase-2, saturated steam (24 g/s) is injected in upward direction below the
centre of the inner cylinder from a nozzle of 138 mm diameter at
an elevation of 1.8 m. The average injection temperature during
phase-2 is 108 C. Detailed injection rates and injection temperatures during phase-1 and phase-2 of the HM-2 test are shown in
Fig. 2. Physical phenomena like convection, turbulent mixing, condensation, heat transfer and distribution of gasses into different
compartmens are simulated in these phases of the HM-2 test. These
phenomena can be very relevant in the course of a severe accident.
3. CFD model
3.1. Introduction
The transient calculations are performed with FLUENT, a commercial general-purpose CFD package supplied by ANSYS Inc.
In all calculations, one half of the THAI vessel is modeled,
assuming symmetry across the vertical plane through the vessel axis. Since isotropic Reynolds averaged Navier-Stokes (RANS)

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

163

Fig. 1. THAI vessel conguration during the HM tests. On the left a three dimensional representation of the THAI vessel with internal structures. On the right a two-dimensional
representation showing the compartments and injection points.

turbulence models are applied, it is expected that this geometrical simplication have no effect as has been veried by Royl
et al. (2009). The modeled three-dimensional geometry is shown
in Fig. 3.
The solid walls and solid internal structures of the THAI vessel are modeled to take into account the effect of heat conduction
and heat capacity. Geometrical information and thermal properties
of the solids are taken from Fischer (2004). The uid in the vessel is modeled as a composition and temperature dependent ideal

Fig. 2. Injection rates and injection temperatures during phase-1 (top) and phase-2
(bottom) of the THAI HM-2 test (Kanzleiter and Fischer, 2008).

Fig. 3. Three-dimensional geometry applied in the CFD calculations; one symmetrical half of the THAI vessel.

164

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

Table 1
General features of the CFD model.

Table 2
Initial conditions for the CFD calculations.

Code version

FLUENT 6.3

Pressure

1.008 bar

Solver
Formulation
Turbulence approach
Pressure interpolation scheme
Pressure correction scheme
Spatial discretization
Temporal discretization
Geometry
Walls
Near-wall treatment
Conduction
Fluid properties

Pressure-based segregated
Transient
RANS
Body-force weighted
PISO
2nd Order upwind
2nd Order implicit
3-Dimensional, half vessel
No slip
Enhanced wall treatment
3-dimensional
Composition and temperature
dependent ideal gas
User-dened function, NRG
condensation model
Increases from 0.1 s (at the start of
each phase) to 0.2 s (after 20 s from the
start of each phase)

Temperature

Linear temperature increase from


18.6 C at the bottom to 23.3 C at
the top of the THAI vessel.
99 vol% nitrogen (1 vol% oxygen is
neglected) 1 vol% steam
u, v, w = 0 m/s (uid initially at rest)
k, = 106 (low turbulence
assumed)

Condensation
Time step size

gas mixture of its constituent components (nitrogen, hydrogen


and steam). The temperature dependence of specic heat, thermal
conductivity and viscosity is implemented by means of a piecewiselinear approach for each of the individual gas components. The
temperature dependent data for nitrogen, hydrogen and steam are
obtained from Lemmon et al. (2007). Also the diffusion coefcients
of the components depend on gas composition and temperature.
The effect of the 1 vol% oxygen initially present is neglected.
In general, the guidelines given in the FLUENT 6.3 Users Guide
(2006) and by ERCOFTAC (2000) are followed for setting up the
CFD model. An overview of the general features of the applied CFD
model is given in Table 1. A condensation model was developed
and used by NRG, since there are no models for wall and bulk condensation available in FLUENT. These condensation processes are
incorporated in the CFD model by means of user-dened functions
(UDF) referred to as the NRG Condensation Model. The NRG Condensation Model, the initial and boundary conditions as well as
the applied meshes and turbulence models are described in the
following subsections.
3.2. NRG condensation model
The implemented NRG Condensation Model takes into account
the following processes:
1.
2.
3.
4.

Bulk condensation and evaporation;


Wall condensation;
Deposition of bulk condensate on walls;
Rainout of bulk condensate.

The condensation/evaporation process in the bulk and at the


wall is modeled by the reaction;
kr

H2 O (g) H2 O (l) + heat,


where the reaction rate kr is controlled by the vapor pressure. Evaporation of condensate from the walls is not expected and not taken
into account. Furthermore, the water condensate that is deposited
on the walls or rains out from the bulk is not treated in the model.
The effect of this water condensate on for instance the ow, heat
transfer and condensation is thus neglected. The NRG condensation
model is described in more detail by Houkema et al. (2008) and has
been employed successfully in the SARNET Condensation Benchmark (Ambrosini et al., 2007) and International Standard Problem
ISP-47 on Containment Thermalhydraulics (Allelein et al., 2007).

Composition
Velocity
Turbulence

3.3. Initial and boundary conditions


The initial conditions for the THAI HM-2 test are specied by
Kanzleiter and Fischer (2008) and are adopted as starting point
of phase-1 in the calculations. The applied initial conditions are
listed in Table 2, where it must be noted that the linear temperature
increase is applied for the solid walls as well.
Hydrogen and steam injection is modeled with mass-ow inlet
boundary conditions. The injection rates and injection temperatures during phase-1 and phase-2 of the HM-2 test are specied by
Kanzleiter and Fischer (2008) and are shown in Fig. 2. These time
dependent boundary conditions are prescribed in tabular form in
the FLUENT CFD code. The turbulence quantities at the inlets are
specied in terms of turbulence intensity (I) and hydraulic diameter
(Dh ).
No-slip boundary conditions are imposed on the solid walls
and solid structures, using the enhanced wall treatment (EWT)
approach in FLUENT to model the ow near the walls. The EWT
is a near-wall modeling method that combines a two-layer model
with (enhanced) wall functions. If the near-wall mesh is ne enough
(typically y+ 1), the EWT will automatically resolve the laminar
sublayer. In all other cases, the EWT will automatically make use
of wall-functions. A similar approach is followed for modeling the
near-wall heat and species transport. The vessels outer wall is
insulated and assumed adiabatic. All other walls are modeled as
uid-solid interfaces with conjugate heat-transfer. Heat transfer
by means of radiation is neglected. Condensation takes place on all
the walls that are in contact with the gas mixture inside the vessel.
3.4. Computational mesh
Four different computational meshes are constructed in order to
study the effect of mesh resolution and near-wall treatment (twolayer model or wall-functions). The four meshes are constructed
in a similar way. Table 3 presents the characteristics of the four
meshes. Fig. 4 shows the standard mesh. The solid regions are
lled with hexahedral cells. The uid regions consist of a hybrid
mesh with tetrahedral and hexahedral cells. The vessels base
region is mostly lled with unstructured tetrahedral mesh cells.
The annulus, cylinder and dome region consist of structured and
unstructured hexahedral cells. The mesh is rened towards the
walls and solid structures in order to resolve the ow and physical
phenomena near the wall in more detail. The mesh at and above
the inlets is rened in order to resolve the small inow area and
the injection jet in more detail.
The characteristics of the four meshes are listed in Table 3. In
the y+ = 1 mesh, the typical cell size is 0.25 mm near the walls and
30 mm 50 mm in the bulk. In the standard, coarse and ne mesh
the typical cell size is 15 mm near the walls and 30 mm 50 mm,
45 mm 75 mm and 20 mm 30 mm in the bulk, respectively.
Fig. 5 compares the four meshes at a section near the inner cylinder.
Depending on the mesh resolution and the ow properties near
the wall, the EWT approach in FLUENT makes use of wall functions or resolves the viscous boundary layer at the wall. The small

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

165

Table 3
Constructed computational meshes.
Mesh

y+ = 1 mesh

Standard mesh

Coarse mesh

Fine mesh

Total number of cells


Number of uid cells
Typical cell size in the bulk (radial vertical direction)
Typical cell size at the wall
Typical y+
Near-wall treatment

763.905
671.225
30 mm 50 mm
0.25 mm
1
Two-layer model

543.438
453.158
30 mm 50 mm
15 mm
520
Wall functions

175.069
139.160
45 mm 75 mm
15 mm
520
Wall functions

2562.528
2197.401
20 mm 30 mm
15 mm
520
Wall functions

Table 4
Considered turbulence settings.
Case

Turbulence
model

Buoyancy effect included in

1 (reference case)
2
3
4
5
6

SKE
SKE
SKW
SKW
SSTKW
SSTKW

k and ( (full buoyancy option)


k (by default)
None (by default)
k (by UDF)
None (by default)
k (by UDF)

near-wall cells in the y+ = 1 mesh make it possible to resolve the


viscous boundary layer near the walls (y+ 1). Wall functions will
be applied in most near-wall regions for the standard, coarse and
ne mesh (y+ > 1).
3.5. Turbulence model

Fig. 4. Front view of the standard mesh on the vessels symmetry plane (left) and
top view of the standard mesh on the horizontal cross section through the hydrogen
inlet at y = 4.8 m (right). The mesh at the hydrogen inlet is shown in more detail for
Sections 1 and 3. The inow area is lled/colored. The mesh resolution near the wall
of the inner cylinder (Section 2) is shown in more detail for the four different meshes
in Fig. 5.

In general, the standard k- turbulence model (SKE) with full


buoyancy effects and default turbulent constants is utilized for the
CFD analyses in this paper (i.e. the reference case). In order to study
the effect of the turbulence model and the buoyancy effects, the calculation for phase-2 of the HM-2 test is repeated using the standard
k- (SKW) and the SST k- turbulence model (SSTKW) with and
without taking into account the effect of buoyancy on turbulence.
An overview of the considered cases is given in Table 4. This sensitivity study on turbulence settings is performed on the coarse mesh.
It will be demonstrated in the next chapter that the resolution of
the coarse mesh sufces to capture the relevant ow phenomena
in the HM-2 test.
The k- models as well as the k- models belong to the class
of two-equation RANS turbulence models. In the k- models turbulence is modeled with the transport equation for the turbulent
kinetic energy (k) and its dissipation rate (). In the k- models turbulence is modeled with the transport equation for the turbulent
kinetic energy (k) and the specic dissipation rate (). The standard
k- model in FLUENT is based on the Wilcox k- model. The SST
k- model is developed by Menter (1994) and combines the best of
the k- and k- formulations, blending the robust and accurate k-
formulation in the near-wall region with the reliable k- formulation in the bulk region. A detailed description of the k- and k-
turbulence models is given in the FLUENT 6.3 Users Guide (2006).
In a non-zero gravity eld, buoyancy forces can suppress or promote turbulence in the presence of density gradients. Buoyancy
tends to suppress turbulence at a stable stratication and buoyancy
promotes turbulence at an unstable stratication. The production
(or dissipation) of turbulence by buoyancy can be incorporated in
the k- and k- turbulence models by adding the source term Gb to
the transport equation for k. The source term Gb is dened as
Gb = gy

Fig. 5. Top view of the four constructed meshes at the wall of the inner cylinder
(Section 2 in Fig. 4).

t 
,
Prt y

where y is the vertical direction, gy the gravitational acceleration


in the y-direction, t the turbulent viscosity,  the density and Prt
the turbulent Prandtl number. In FLUENT, Gb is only included by
default in the k-equation of all the k- models. It is however possible

166

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

to incorporate Gb in the k-equation of the k- models as well by


means of a so called user-dened function (UDF).
Since the effect of buoyancy on the turbulent dissipation rate
is not well understood, by default this effect is not included in the
transport equation for in FLUENT (FLUENT, 2006). A certain effect
of buoyancy on can be included in the k- models of FLUENT by
activating the full buoyancy option. The effect of buoyancy on
can be observed by comparison of cases 1 and 2 in Table 4. The effect
of buoyancy is not included in the -equation of the k- models in
FLUENT and is therefore not considered here.
4. Results
In this chapter, the results of the CFD analyses are presented
and compared to the HM-2 experiment performed by Becker Technologies (EXPBT). The results of the sensitivity analyses on mesh
resolution and turbulence models/settings are presented here as
well. This chapter is divided into the following sections:
Section 4.1: comparison of the experimental and CFD results for
phase-1.
Section 4.2: comparison of the experimental and CFD results for
phase-2.
Section 4.3: effect of mesh resolution and near-wall treatment.
Section 4.4: effect of turbulence model and buoyancy effects.
The CFD results presented in Sections 4.1 and 4.2 are obtained
with the standard k- turbulence model (SKE) with full buoyancy
effects on the y+ = 1 mesh.
4.1. Phase-1 results (04300 s)
In phase-1 of the HM-2 test, a total amount of 1.24 kg hydrogen and 1 kg saturated steam is injected into the THAI vessel from
a vertical pipe in the annulus. Since the THAI vessel is closed,
the hydrogen and steam content will increase, and therewith the
pressure as well. The measured and calculated evolution of the
atmospheric pressure and hydrogen mass in the vessel during
phase-1 is shown in Fig. 6. The CFD calculation shows an accurate
mass balance and a negligible error of less than 0.2% in the amount
of hydrogen in the system. The calculation shows, however, a slight
over-prediction of the vessel pressure (about 0.02 bars at the end
of phase-1).
The injected gas mixture of hydrogen and steam has a relatively
low density and forms a stable stratied hydrogen-rich gas layer in
the upper part of the THAI vessel. Fig. 7 shows the hydrogen distribution at the end of phase-1 on a vertical line through the annulus.
The experiment and calculation show a similar hydrogen distribution in the THAI vessel. Initially the vessel contains no hydrogen. At
the end of phase-1, a hydrogen-rich gas layer is formed in the upper
half of the vessel, while the hydrogen concentration remains below
1% in the lower half of the vessel. A strong gradient in hydrogen concentration is observed at an elevation of 45 m. The inset in Fig. 7 is
a contour plot of the predicted hydrogen concentration on the symmetry plane at 4300 s. This contour clearly shows that the hydrogen
accumulates and is well mixed in the upper half of the vessel. The
measured and predicted values of the hydrogen concentration in
the hydrogen-rich gas layer differ slightly. In the experiment, concentrations up to 37 vol% are found in the upper part of the vessel,
against concentrations up to 35.5 vol% in the calculation.
During phase-1, a mixture of hydrogen and saturated steam is
injected at an average temperature of 45 C. The temperature of
the gas and the solid structures in the vessel is lower, which causes
about 50% of the injected steam to condensate in the CFD calculation. Condensation of steam lowers the pressure in the vessel and

Fig. 6. Comparison of the measured (EXPBT) and predicted (CFD) atmospheric pressure (top) and amount of hydrogen (bottom) in the vessel during phase-1.

increases the concentration of the other species in the gas mixture.


Therefore, under-prediction of condensation could be a possible
reason for the observed deviation in pressure and hydrogen concentration between the CFD calculation and the experiment. This

Fig. 7. Comparison of the measured (EXPBT) and predicted (CFD) hydrogen concentration at the end of phase-1 on a vertical line running through the annulus from
the bottom to the top of the THAI vessel (see dashed line in inset). The inset shows
a contour plot of the predicted hydrogen concentration on the symmetry plane at
the end of phase-1.

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

167

effect of condensation and its importance in the THAI HM-2 test


is also demonstrated by Royl et al. (2009). Under-prediction of
condensation can have different causes, for instance the implemented condensation model, the imposed initial conditions (gas
composition and temperature distribution) and/or the modeled
heat transfer to the solids. The inuence of the heat losses to the
solids for the THAI HM-2 test is also considered by Schwarz et al.
(2010) and Bentaib and Bleyer (2011). This heat transfer effect can
be understood by observing the temperatures of the gas mixture
in the vessel. The predicted gas temperature in the hydrogen-rich
layer in the upper part of the vessel is about 1 C higher than measured in the experiment, which indicates that heat transfer from the
gas to the solids is slightly under-predicted. At higher temperature,
less steam will condensate from the humid, hydrogen-rich layer.
At the end of phase-1, the predicted steam concentration in the
hydrogen-rich layer is still around 2 vol%. Since limited experimental data is available on the steam distribution in the vessel, a direct
comparison to the measured steam concentrations cannot be made
and no conclusive explanation can be given for the observed differences in pressure and hydrogen concentration between calculation
and experiment.
4.2. Phase-2 results (43006860 s)
In phase-2 of the HM-2 test, saturated steam at relatively high
temperature is injected into the THAI vessel from a nozzle below
the centre of the inner cylinder. The injected steam rst clears the
inner cylinder and then starts to dissolve the stable hydrogen-rich
stratication in the vessel dome from below. These two consecutive
processes divide phase-2 into a stagnation and a natural circulation period. The point in time where this natural circulation
period starts is referred to as the onset of natural circulation. During phase-2 the amount of steam in the vessel, as predicted by the
CFD calculation, increases from 1 kg at 4300 s to 6 kg at 6500 s. At
6500 s the amount of steam injected is 53 kg, which means that
over 90% of the injected steam condensates. Condensation of steam
is, thus, very important during phase-2.
The predicted atmospheric pressure and ow velocity above
the centre of the inner cylinder during phase-2 are compared to
the experimental results in Fig. 8. A vertical dashed line is drawn
in the gures at 4300 s and at 4820 s, which indicate the start of
phase-2 and the onset of natural circulation as determined from
experiment, respectively. The period from 4300 s to 4820 s is the
stagnation period. The period after 4820 s is the natural circulation
period.
Fig. 8 shows that the measured and predicted evolution of pressure follow the same trend. The over-prediction in pressure at
the end of phase-1 remains more or less constant during phase2. Fig. 8 also compares the vertical ow velocity above the centre
of the inner cylinder during phase-2. The measured and predicted
ow velocity agree well on average, and uctuations have about
the same amplitude. The start of the natural circulation period is
evident by the sudden signicant increase in ow velocity around
4800 s. When the onset of natural circulation is determined as the
time when the ow velocity above the inner cylinder stays above
0.15 m/s, the onset found from experiment and CFD is 4820 s and
4720 s, respectively. The earlier onset of natural circulation in the
CFD calculation are expected to be the result of the slightly different starting conditions for phase-2 compared to the experiment. A
second increase in ow velocity is observed around 6000 s when
the hydrogen-rich gas layer is almost completely dissolved. The
velocity drops to zero when the injection of steam is stopped.
Fig. 9 compares the hydrogen concentration during phase-2 at
four different locations in the vessel. Each location is in a different compartment of the vessel (base, annulus, cylinder and dome)
as shown in the insets in Fig. 9. Fig. 10 shows snapshots of the

Fig. 8. Atmospheric pressure (top) and vertical ow velocity above the centre of
the inner cylinder (bottom) during phase-2. The experiment ends at 6860 s, the
calculation runs up to 6500 s.

predicted hydrogen distribution and ow eld on the symmetry


plane at 4300 s, 4500 s and 5000 s. These snapshots are helpful to
illustrate and understand the processes during phase-2. As shown
in Figs. 9 and 10, the hydrogen concentration in the upper part
of the vessel is uniformly high at the start of phase-2, while it is

Fig. 9. Hydrogen concentrations in the vessel during phase-2 as measured (symbols)


and computed (solid lines) for the THAI HM-2 test.

168

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

Table 5
Effect of mesh resolution.
Mesh

y+ = 1 mesh

Standard mesh

Coarse mesh

Fine mesh

Pressure at 4300 s/5000 s (bar)


Max. H2 concentration at 4300 s (vol%)
Amount of H2 at 4300 s (kg)
Onset of natural circulation (s)

1.281/1.461
35.33
1.232
4720

1.277/1.436
35.61
1.243
4780

1.278/1.437
35.46
1.242
4790

1.276/1.449
35.56
1.241
4770

uniformly low in the lower part of the vessel. An interface is present


in-between these two regions where the hydrogen concentration
varies.
In the stagnation period, the upper end of the inner cylinder
is blocked by the light gas cloud in the vessel dome and most of
the injected hot steam remains in the inner cylinder, quickly displacing and diluting the gas mixture in the inner cylinder. This
process can be observed in Fig. 9 by the sharp decrease of hydrogen concentration in the cylinder just after the start of phase-2. The
snapshot at 4500 s in Fig. 10 clearly shows that the steam circulates
in the inner cylinder and displaces the hydrogen-rich gas mixture
initially present. It can also be observed from this snapshot that
small portions of the steam-rich gas mixture spill over the upper
edge of the inner cylinder and ow down along the outside of the
cylinder wall. This process mixes the atmosphere in the annulus
in downward direction as shown in Fig. 9 by the gradual decrease
of hydrogen concentration in the annulus. The concentration in the
upper (dome) and bottom (base) part of the vessel stay almost unaffected during the stagnation period. All these trends are predicted
well by the CFD analysis.
Once the atmosphere in the annulus is mixed down to the
lower edge of the inner cylinder, the steam-rich atmosphere in the

annulus and the inner cylinder connect, forming a closed loop. This
is the start of the natural circulation period where a circulating ow
sets in that, driven by the steam injection, moves up through the
inner cylinder and down through the annulus as illustrated by the
snapshot at 5000 s in Fig. 10. The atmosphere in the inner cylinder
and annulus is quickly mixed after the onset of this convection loop,
averaging out the hydrogen concentrations in these regions. This
process can be clearly observed in Fig. 9 by the sharp rise of hydrogen concentration in the cylinder around 4820 s for the experiment
and 4720 s for the CFD analysis. The convection loop through and
out of the inner cylinder starts to dissolve the stratied hydrogenrich gas layer from below and homogenize the atmosphere in the
vessel. This is observed in Fig. 9 by the drop of hydrogen concentration in the dome of the vessel around 5200 s. Again, these trends
are predicted well with CFD.
4.3. The effect of mesh resolution and near-wall treatment
In this section, the effect of mesh resolution on the dissolution
process during phase-2 is studied using the four meshes described
in Section 3.4. This mesh sensitivity study is performed with the
standard k- turbulence model with EWT and full buoyancy. In all

Fig. 10. Snapshots of the predicted hydrogen distribution and ow eld on the symmetry plane of the THAI vessel at the start of phase-2 (4300 s), in the stagnation period
(4500 s) and in the natural circulation period (5000 s).

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

169

Fig. 11. Comparison of the measured hydrogen concentrations in the vessel (symbols) with those computed on the y+ = 1 mesh (solid lines) and standard mesh (dotted
lines).

calculations the convergence was in the order of 105 for ow and


species, and in the order of 109 for energy.
The y+ = 1 and standard mesh have the same mesh resolution in
the bulk and a different mesh resolution near the wall. To assess
the effect of near-wall treatment, the results obtained on the y+ = 1
and standard mesh are compared qualitatively in Fig. 11 and quantitatively in Table 5. Fig. 11 shows again the development of the
hydrogen concentration over time at the four locations in the vessels base, annulus, cylinder and dome compartment. Table 5 shows
values of pressure, hydrogen concentration, mass conservation and
the time of onset of natural circulation for the different meshes. The
results obtained on the standard mesh with wall functions and on
the y + =1 mesh with the two-layer model are very similar. The overall mixing process during phase-2 is slightly slower for the standard
mesh. Nevertheless, the results agree well and are both close to the
experimental results, meaning that wall functions are applicable to
model the THAI HM-2 test.
The standard, coarse and ne mesh have the same mesh resolution near the wall and a different mesh resolution in the
bulk (see Table 3). The effect of bulk mesh resolution can be
observed in Fig. 12 and Table 5, where the results obtained on the

Fig. 13. Hydrogen concentrations in the vessel atmosphere during phase-2 as measured (EXPBT) and computed with different turbulence models and buoyancy terms.

Fig. 12. Comparison of the hydrogen concentrations in the vessel as computed on


the standard mesh (solid line), coarse mesh (broken line), and ne mesh (dotted
line).

170

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

Fig. 14. Snapshots of the hydrogen distribution on the symmetry plane of the THAI vessel at 4500 s as predicted with the standard k- turbulence model (by default with
Gb ), the SST k- turbulence model with Gb and the SST k- turbulence model without Gb .

standard, coarse and ne mesh are compared. The differences are


small, which shows that a cell size of 45 mm 75 mm in the bulk is
sufciently small to model the THAI HM-2 test.
4.4. The effect of turbulence model
In this section, the effects of the turbulence model and buoyancy
terms in the turbulence models on the prediction of the dissolution process during phase-2 are studied for the settings given in
Table 4. All analyses are performed on the coarse mesh. Apart from
the turbulence settings, the analyses are performed with the same
settings. In order to assure identical starting conditions for phase2, all calculations start at 4300 s on the phase-1 solution obtained
with the standard k- model with full buoyancy on the coarse mesh
(case 1 in Table 4).
Fig. 13 compares the computed hydrogen concentrations in the
vessel atmosphere during phase-2 for the different turbulence settings listed in Table 4. The stagnation period and onset of natural
circulation is predicted reasonably well in all calculations. However, compared to the experiment, the dissolution of the hydrogen
rich cloud sets in too early and the dissolution rate is higher for
the standard k- and SST k- models without Gb in k, as implemented by default in FLUENT. The prediction of both these k-
models improves when Gb is included in the transport equation
of the turbulent kinetic energy k. Although the standard k- model
with Gb in k still predicts a slightly higher dissolution rate of the
hydrogen rich cloud compared to the experiment, the results of
the SST k- model with Gb agree well with experiment and are
very similar to the results of the standard k- model with Gb . Apparently, the generation (or dissipation) of turbulent kinetic energy by
buoyancy has a large effect on the stability and dissolution process

of the stratied hydrogen-rich layer. This inuence of buoyancy can


be understood as will be explained in the next paragraph. Including the buoyancy effect in the turbulent dissipation rate of the
standard k- model shows no signicant inuence.
Fig. 14 shows the hydrogen distribution on the symmetry plane
at 4500 s as calculated with the standard k- model with Gb in k and
, the SST k- model with Gb in k and the SST k- model without
Gb . At 4500 s the HM-2 test is in the stagnation period of phase-2.
In the experiment, the injected steam ows into the inner cylinder
during the stagnation phase without affecting the hydrogen concentration in the dome area above the inner cylinder. As shown in
Figs. 13 and 14, this is only true for the calculation with the standard
k- turbulence model and the SST k- model with Gb . The snapshots
in Fig. 14 show that in the calculation of the SST k- model without
Gb , some of the steam owed out of the inner cylinder and mixed
with the hydrogen rich cloud at 4500 s. This proves that it is important to take the effect of buoyancy into account in the turbulence
model by including the Gb term. Turbulence is suppressed by buoyancy (Gb < 0) near the stratied hydrogen rich layer. Without this
damping effect, the stratied layer above the inner cylinder will be
dissolved faster by turbulent mixing. Consequently, the turbulence
models without Gb do not correctly capture the dissolution process
during phase-2.
5. Conclusions
In the present work, the containment model of NRG is further
validated in the context of hydrogen distribution in a containment using the THAI HM-2 test. For this test, the characteristic
phenomena are the development and dissolution of a hydrogenrich gas layer in the upper part of the THAI containment by a

D.C. Visser et al. / Nuclear Engineering and Design 245 (2012) 161171

171

buoyant plume. These phenomena are predicted well by NRGs


containment model developed in FLUENT. Additional sensitivity
analyses on mesh resolution, near-wall treatment, and turbulence
model and turbulence model settings, showed the following:

the authors gratefully acknowledge the funding provided by the


Department of Nuclear Safety, Security and Safeguards (KFD) that
is presently part of the Dutch Ministry of Infrastructure and Environment.

A cell size of 45 mm 75 mm in the bulk is sufciently small to


model the hydrogen distribution and dissolution processes in the
THAI HM-2 test with CFD.
Good CFD results are obtained when wall functions are applied
to model the ow in the near wall region for the HM-2 test.
The effect of buoyancy on the turbulent kinetic energy k is taken
into account by default in the k- turbulence models of FLUENT.
This effect on k is not incorporated by default in the k- models
of FLUENT, but it can be incorporated by user coding.
The prediction of the dissolution process during phase-2 of the
HM-2 test improves when the effect of buoyancy on k is included
in the turbulence models. Without this effect the CFD model predicts a too early start of the dissolution, as well as a too high
dissolution rate of the stratied hydrogen-rich layer.
The results of the standard k- model and SST k- turbulence
model with the effect of buoyancy on k included are very similar and agree well with experiment, whereas the standard k-
turbulence model with the effect of buoyancy on k included still
predicts a higher dissolution rate.
The effect of buoyancy on the turbulent dissipation rate can be
included in the k- models of FLUENT by activating the full buoyancy option. The buoyancy effect on is subject to discussion in
literature. For the HM-2 test, there was no signicant impact on
the results obtained with CFD.

References

These ndings, combined with the extensive model description as presented in this paper may serve others in improving the
predictive quality of CFD for containment analyses. Furthermore,
these ndings and settings can develop into a set of best practice
guidelines for the application of CFD for containment analyses. To
this end, more experimental tests should be analyzed in the future
to conrm the general applicability of the presented containment
model and model settings. Work in this area is currently ongoing
at NRG.
Acknowledgments
The authors thank all signatories and participants to the OECDNEA THAI project. Special thanks are expressed to the people
involved in the experiments and documentation. In addition,

Allelein, H.J., et al., 2007. International Standard Problem ISP-47 on Containment


Thermalhydraulics, Final Report. NEA/CSNI, OECD, Paris.
Ambrosini, W., et al., 2007. Results of the SARNET Condensation Benchmark No. 0.
DIMNP 006(2007). University of Pisa.
Andreani, M., Paladino, D., 2010. Simulation of gas mixing and transport in a multicompartment geometry using the GOTHIC containment code and relatively
coarse meshes. Nucl. Eng. Des. 240, 15061527.
Bentaib, A., Bleyer, A., 2011. ASTEC validation on OECD/THAI HM-2. In: Proceedings
of ICAPP11, Nice, France, May 25, Paper 11258.
ERCOFTAC, 2000. Best Practice Guidelines, www.ercoftac.org.
Fischer, K., 2004. International Standard Problem ISP-47 on Containment ThermalHydraulics, Step 2: ThAI, Volume 1: Specication Report. Becker Technologies
GmbH, Eschborn, Report No. BF-R 70031-1 Rev. 4.
FLUENT 6.3 Users Guide, 2006. FLUENT Inc.
GASFLOW, 2011. http://hycodes.net/gasow.
Gupta, S., Kanzleiter, T., Fischer, K., Langer, G., Poss, G., 2010. Interaction of a stratied light gas layer with a buoyant jet in containment: hydrogen/helium material
scaling. In: Proceedings of ICAPP10, San Diego, CA, USA, June 1317, Paper
10205.
Heitsch, M., Baraldi, D., Wilkening, H., 2010. Simulation of containment jet ows
including condensation. Nucl. Eng. Des. 240, 21762184.
Houkema, M., Siccama, N.B., Lycklama Nijeholt, J.A., Komen, E.M.J., 2008. Validation
of the CFX4 CFD code for containment thermal-hydraulics. Nucl. Eng. Des. 238,
590599.
Kanzleiter, T., Fischer, K., 2008. Quick Look Report Helium/Hydrogen Material Scaling Test HM-2. Becker Technologies GmbH, Eschborn, Germany, OECD-NEA THAI
Project Report No. 150 1326 HM-2 QLR Rev. 3.
Kudriakov, S., Dabbene, F., Studer, E., Beccantini, A., Magnaud, J.P., Paillre, H., Bentaib, A., Bleyer, A., Malet, J., Porcheron, E., Caroli, C., 2008. The TONUS CFD code
for hydrogen risk analysis, physical models, numerical schemes and validation
matrix. Nucl. Eng. Des. 238, 551565.
Lemmon, E.W., Huber, M.L., McLinden, M.O., 2007. NIST Standard Reference
Database 23: Reference Fluid Thermodynamic and Transport PropertiesREFPROP, Version 8. 0. National Institute of Standards and Technology, Standard
Reference Data Program, Gaithersburg.
Menter, F.R., 1994. Two-equation Eddy-viscosity turbulence models for engineering
applications. AIAA J. 32 (8), 1598.
Prabhudharwadkar, D.M., Iyer, K.N., Mohan, N., Bajaj, S.S., Markandeya, S.G., 2011.
Simulation of hydrogen distribution in an Indian nuclear reactor containment.
Nucl. Eng. Des. 241, 832842.
Royl, P., Travis, J.R., Breitung, W., Kim, J., Kanzleiter, T., Schwarz, S., 2009. GASFLOW
analysis of steam/hydrogen mixing with nitrogen in the OECD-NEA THAI HM-2
benchmark. In: Proc. of NURETH-13, Kanazawa, Japan, September 27October
2, Paper N13P1412.
Schwarz, S., Fischer, K., Bentaib, A., Burkhardt, J., Lee, J.-J., Duspiva, J., Visser, D.,
Kyttala, J., Royl, P., Kim, J., Kostka, P., Liang, R., 2010. Benchmark on hydrogen
distribution in a containment based on the OECD-NEA THAI HM-2 experiment.
Nucl. Technol. 175, 594603.
SOAR, 1999. SOAR on Containment Thermalhydraulics and Hydrogen Distribution,
NEA/CSNI/R(1999)16.